Re: regexp-opt GNU "sync"?

On Tue, 2003-10-14 at 14:49, Stephen J. Turnbull wrote:
> >>>>> "Hrvoje" == Hrvoje Niksic <[email protected]> writes:
> 
>     Hrvoje> Unless I'm mistaken, regexp-opt is maintained by an FSF
>     Hrvoje> Emacs person.  I don't see that we have a choice but to
>     Hrvoje> sync it.
> 
> Good, let's do it, and maintain the minimum amount of backward
> compatibility.

Ok.  Will start (slowly) hacking it.

>     Hrvoje> Didn't they break compatibility with themselves as well,
>     Hrvoje> or is it yet another case of importing an alpha FSF
>     Hrvoje> feature into a production XEmacs and then watching them
>     Hrvoje> change the API?
> 
> This was work done by Karl Hegbloom in like 1997.  Didn't we get shy
> groups well before GNU did?  I wouldn't be surprised if GNU simply
> came up with a better API independently, while nobody was looking.

That's what I think has happened.

I wouldn't consider the current API as "alpha", apart from the 'words
addition (which is already synced to XEmacs) the API has stayed
consistent with the original *documentation* [1] written in 1997 [2],
they nowadays Just Do The Right Thing including use of shy groups where
appropriate.
http://savannah.gnu.org/cgi-bin/viewcvs/emacs/emacs/lisp/emacs-lisp/regexp-opt.el

[1] "at least" in "enclosed by at least one regexp grouping construct"
    allows for changing the internal groups between shy/capturing 
    conveniently without breaking the API contract :)

[2] For some reason in recent versions some of the documentation has
    been "fixed" by converting docstrings into comments.  Beats me.